Error detection/correction and fault detection/recovery – Pulse or data error handling – Error detection for synchronization control
Reexamination Certificate
2000-03-07
2003-10-21
Decady, Albert (Department: 2133)
Error detection/correction and fault detection/recovery
Pulse or data error handling
Error detection for synchronization control
C714S704000
Reexamination Certificate
active
06637006
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates to digital data transmission and in particular it relates to digital data decoders. The following description is based on the GSM cellular communications system for which the invention is of particular utility. It will be apparent to those skilled in the art, however that the invention may be applied to other systems of digital data transmission.
2. Description of the Related Art
Reference is made to U.S. Pat. Nos. 5,598,506 and 5,596,678 to Wigren et al., U.S. Pat. No. 5,557,639 to Heikkila and “Mobile Radio Communications” published by John Wiley & Sons, Raymond Steele (Ed.) for a description of the prior art and technological background. The following abbreviations are used therein:
GSM : Global System for Mobile communications, (formerly Group Special Mobile)
TCH/FS : Traffic channel full rate
TCH/HS : Traffic channel half rate
CRC : Cyclic redundancy check
PBER : Pseudo bit error rate
The GSM cellular communications system may use the Half Rate speech codec. The Half Rate speech codec encodes sixteen, 8 kHz samples into 112 bits containing 18 parameters. These 112 bits are divided into two groups based on their subjective importance to speech quality. The 17 least important bits are known as class II bits and are unprotected.
Corruption of these class II bits has minimal audible effect on speech quality. The most important 95 bits are known as class I bits and are protected by additional convolutional coding. The class I bits are further subdivided into Ia and Ib, such that the most significant 22 bits (Ia) are additionally protected by a 3 bit cyclic redundancy check (CRC). All of the class I bits and class Ia CRC check bits are protected by a ⅓ rate convolutional code of constraint length K=7, which is punctured to reduce the number of transmitted encoded bits by one third before transmission.
In order to prevent any audio artifacts after speech transmission and decoding, any frame erasure and concealment mechanism must detect all frames with any class Ia errors and frames with more than a certain number of class Ib errors, as precisely and efficiently as possible for all propagation channel types. For the GSM Full Rate and GSM Enhanced Full Rate speech codecs this is achieved by marking such frames as bad to the speech codecs using an algorithm referred to as the frame erasure algorithm.
In this respect the Half Rate Speech codec differs from the other GSM specified codecs because in addition to a mechanism for marking a received speech frame as bad, it requires are extra mechanism for marking a received frame as unreliable and utilises an algorithm known as the “unreliable frame erasure algorithm”.
When a received speech frame is marked as bad the speech codec erases the frame and applies a concealment algorithm in order to mask the effects of the bad frame from the user. When a received speech frame is marked as unreliable the speech codec performs its own validation check on the speech frame. If the validation check fails then the speech codec erases the failing speech frame. If the validation check passes then the speech frame is considered to be good and is used by the codec. This internal test performed by the speech codec is based on exploiting known properties of human speech, namely that the energy in speech varies slowly and thus it is unlikely that abrupt changes will be experienced.
The internal test is implemented by calculating the difference of an energy metric between the last good received speech frame and the current unreliable frame. If this energy difference is greater than some predetermined threshold then the frame is deemed bad and erased.
Thus it is apparent that two different algorithms are required by the Half Rate speech codec, one for detecting bad frames and the other for detecting unreliable frames. The conventional bad frame detection algorithm is composed of a 3 bit CRC check and a PBER threshold test.
The CRC check is made on the received speech frame by recalculating the CRC for the 22 class Ia bits and comparing this value to the received value. If they are different the CRC check is deemed to have failed. The PBER is calculated by re-encoding and puncturing the decoded received class I bits, comparing them bit by bit with the original received class I bits and counting the number of differences. If the specified PBER (P
bad
) threshold has been exceeded then the PBER threshold test is deemed to have failed. If either of these two tests fail then the received speech frame is marked as bad.
The conventional unreliable frame detection algorithm also makes use of a PBER threshold test for which the PBER is calculated in the manner described above, except that the threshold P
unrel
is less than P
bad
. Frames which have a calculated PBER higher than P
bad
are marked as bad. Frames with a PBER greater than P
unrel
but lower than P
bad
are marked as unreliable. Frames with a PBER less than P
unrel
are by definition good frames. A frame marked as bad is automatically considered unreliable whereas an unreliable frame is not necessarily bad.
The method of judging unreliable frames described above does not work as efficiently as is possible because the method is based on the assumption that the calculated PBER is highly correlated with the true number of bits in error. This assumption does not hold when convolutional decoding breakdown begins to occur and unreliable frames are not marked as such.
SUMMARY OF THE INVENTION
An object of the present invention is to improve unreliable frame detection performance for speech channels such as the Half Rate GSM speech channel, to provide better audio performance.
According to the invention a frame which is decoded by a first decoder is compared with a frame of which bits are reversed in a reversed direction by a first time reverse unit and decoded by a second decoder and bits are reversed in a time reversed direction by a second time reverse unit. If any difference exist between the two frames then received frame is judged as unreliable.
The above and other objects, features, and advantages of the present invention will become apparent from the following description when taken in conjunction with the accompanying drawings which illustrate a preferred embodiment of the present invention by way of example.
REFERENCES:
patent: 4870645 (1989-09-01), Herron
patent: 5557639 (1996-09-01), Heikkila et al.
patent: 5596678 (1997-01-01), Wigren et al.
patent: 5598506 (1997-01-01), Wigren et al.
patent: 5666370 (1997-09-01), Ganesan et al.
patent: 5850405 (1998-12-01), Wimmer et al.
patent: 6092230 (2000-07-01), Wood et al.
patent: 6397358 (2002-05-01), Burton et al.
patent: 0 643 493 (1995-03-01), None
patent: 0 648 032 (1995-04-01), None
patent: 0 805 572 (1997-11-01), None
patent: 2 312 359 (1997-10-01), None
patent: 97/14235 (1997-04-01), None
patent: 97/27686 (1997-07-01), None
Chaudry Mujtaba K.
De'cady Albert
NEC Corporation
LandOfFree
Digital data decoder with improved unreliable frame... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Digital data decoder with improved unreliable frame..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Digital data decoder with improved unreliable frame... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3169371